The claim nobody can check
"We anonymize data." Every ad-tech vendor says it. Almost none of them publish anything you could use to verify it: no list of quasi-identifiers, no k threshold, no linkage test results. Names and emails get stripped, the dataset gets called anonymous, and everyone moves on.
The problem is that names were never the hard part. The hard part is the boring fields: postal district, device fingerprint, browsing hours. Combine enough of them and rows become unique to one person, no name required.
A claim you cannot test is not a safeguard. It is marketing.
The test
The test is simple to state: take a dataset someone calls anonymous and try to re-identify it. The classic move is a linkage attack. Combine quasi-identifiers with outside data and count how many rows become unique to one person.
Two yardsticks make it reproducible. k-anonymity asks how many people share each combination of quasi-identifiers. If the answer is one, the data is not anonymous. l-diversity asks whether sensitive values stay mixed inside each group, so that even linked rows do not expose one person's secret. Same input, same parameters, same answer. Anyone can replicate it, which is exactly what makes it a test instead of a slogan.
What Law 25 changes
Under Quebec's Law 25, consent for non-essential cookies has to be real opt-in: manifest, free, enlightened, and given for specific purposes. Pre-ticked boxes and "continued browsing means consent" do not meet that bar.
Toggle the banner in the demo below and watch what changes. With the banner up and marketing cookies rejected, the jars hold only pseudonymous IDs with nothing to join on, and the profile collapses. Consent is not a popup. It is the difference between a profile and a pile of IDs.
What you are worth
Every jar in the demo carries a fictional bid price: what advertisers would pay per thousand impressions for that slice of a profile. The numbers are invented. The mechanism is not. Real-time bidding broadcasts quasi-identifiers to hundreds of vendors on nearly every page load.
Watch the total move as you toggle the banner. That gap, between the profiled price and the consented price, is the market value of your consent.


